Benchmark Dataset for Training Machine Learning Models to Predict the Pathway Involvement of Metabolites

Huckvale, Erik D.; Powell, Christian D.; Jin, Huan; Moseley, Hunter N. B.

doi:10.3390/metabo13111120

Cited by 7 publications

(13 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Huckvale et al [5] previously generated a dataset of 5,683 entries, each entry containing a vector of atom-color [8] features corresponding to a metabolite. Building off of this dataset of metabolite features, we constructed pathway features via the process in Figure 1.…”

Section: Methodsmentioning

confidence: 99%

“…Table 1 provides characteristics of the individual metabolite feature sets and pathway feature sets before they were paired together. Since we built off of the work of Huckvale et al [5], the number of metabolite entries and features were the same as with their work, the metabolite features having already been de-duplicated. Since the pathway features were derived from the metabolite features, the number of pathway features was initially the same as that of the metabolite features.…”

Section: Generating the Feature Vectorsmentioning

confidence: 99%

“…In this work, we present a single binary classifier for predicting the pathway involvement of metabolites. The dataset provided by the work of Huckvale et al [5] uses an atom coloring technique [8] to represent the sub-structures of molecules, the features of a metabolite being the count of such molecular substructures that are present in the molecule. Every pathway category has certain metabolites associated with it, so a pathway category can be generically represented by aggregations of the features of the metabolites associated with the pathway category.…”

Section: Introductionmentioning

confidence: 99%

“…However, due to the costly and laborious nature of experimentally determining the pathway involvement of metabolites, many of the metabolites in such metabolic knowledge-bases are still not annotated or are only partially annotated to metabolic pathways. For example, as of July 3 rd 2023, a total of 19,119 compounds existed in the KEGG database with only 6,736 of them having annotated pathway involvement [5]. Because of the lack of annotation, several machine learning methods have been proposed to predict the pathway categories that a metabolite is involved in given information about the metabolite’s chemical structure.…”

Section: Introductionmentioning

confidence: 99%

“…Several past publications trained models to predict 11 out of the 12 categories, excluding ‘Chemical structure transformation maps’, likely due to its difficulty to predict. Since some of these publications were proven to be invalid, using a dataset that contained duplicate entries in both the train and test set [7], Huckvale et al developed a new KEGG-derived benchmark dataset for the task of developing models for predicting pathway category involvement for all 12 of the aforementioned categories based on metabolite chemical structure [5]. Huckvale et al also demonstrated a set of binary classification models trained on this benchmark dataset for predicting pathway category involvement.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Predicting The Pathway Involvement Of Metabolites Based on Combined Metabolite and Pathway Features

Huckvale,

Moseley

2024

Preprint

Self Cite

View full text Add to dashboard Cite

A major limitation of most metabolomics datasets is the sparsity of pathway annotations of detected metabolites. It is common for less than half of identified metabolites in these datasets to have known metabolic pathway involvement. Trying to address this limitation, machine learning models have been developed to predict the association of a metabolite with a "pathway category", as defined by one of the metabolic knowledgebases like the Kyoto Encyclopedia of Gene and Genomes. Most of these models are implemented as a single binary classifier specific to a single pathway category, requiring a set of binary classifiers for generating predictions for multiple pathway categories. This single binary classifier per pathway category approach both multiplies the computational resources necessary for training while diluting the positive entries in gold standard datasets needed for training. To address the limitations of training separate classifiers, we propose a generalization of the metabolic pathway prediction problem using a single binary classifier that accepts both features representing a metabolite and features representing a generic pathway category and then predicts whether the given metabolite is involved in the corresponding pathway category. We demonstrate that this metabolite-pathway features-pair approach is not only competitive with the combined performance of training separate binary classifiers, but it outperforms the previous benchmark models.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Generating the Feature Vectorsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Predicting The Pathway Involvement Of Metabolites Based on Combined Metabolite and Pathway Features

Huckvale,

Moseley

2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Current approaches and outstanding challenges of functional annotation of metabolites: a comprehensive review

Nguyen,

et al. 2024

Briefings in Bioinformatics

View full text Add to dashboard Cite

Metabolite profiling is a powerful approach for the clinical diagnosis of complex diseases, ranging from cardiometabolic diseases, cancer, and cognitive disorders to respiratory pathologies and conditions that involve dysregulated metabolism. Because of the importance of systems-level interpretation, many methods have been developed to identify biologically significant pathways using metabolomics data. In this review, we first describe a complete metabolomics workflow (sample preparation, data acquisition, pre-processing, downstream analysis, etc.). We then comprehensively review 24 approaches capable of performing functional analysis, including those that combine metabolomics data with other types of data to investigate the disease-relevant changes at multiple omics layers. We discuss their availability, implementation, capability for pre-processing and quality control, supported omics types, embedded databases, pathway analysis methodologies, and integration techniques. We also provide a rating and evaluation of each software, focusing on their key technique, software accessibility, documentation, and user-friendliness. Following our guideline, life scientists can easily choose a suitable method depending on method rating, available data, input format, and method category. More importantly, we highlight outstanding challenges and potential solutions that need to be addressed by future research. To further assist users in executing the reviewed methods, we provide wrappers of the software packages at https://github.com/tinnlab/metabolite-pathway-review-docker.

show abstract

Predicting the Pathway Involvement of Metabolites in Both Pathway Categories and Individual Pathways

Huckvale,

Moseley

2024

Preprint

View full text Add to dashboard Cite

Metabolism is the network of chemical reactions that sustain cellular life. Parts of this metabolic network are defined as metabolic pathways containing specific biochemical reactions. Products and reactants of these reactions are called metabolites, which are associated with certain human-defined metabolic pathways. Metabolic knowledgebases, such as the Kyoto Encyclopedia of Gene and Genomes (KEGG) contain metabolites, reactions, and pathway annotations; however, such resources are incomplete due to current limits of metabolic knowledge. To fill in missing metabolite pathway annotations, past machine learning models showed some success at predicting KEGG Level 2 pathway category involvement of metabolites based on their chemical structure. Here, we present the first machine learning model to predict metabolite association to more granular KEGG Level 3 metabolic pathways. We used a feature and dataset engineering approach to generate over one million metabolite-pathway entries in the dataset used to train a single binary classifier. This approach produced a mean Matthews correlation coefficient (MCC) of 0.806 ± 0.017 SD across 100 cross-validations iterations. The 172 Level 3 pathways were predicted with an overall MCC of 0.726. Moreover, metabolite association with the 12 Level 2 pathway categories were predicted with an overall MCC of 0.891, representing significant transfer learning from the Level 3 pathway entries. These are the best metabolite-pathway prediction results published so far in the field.

show abstract

Benchmark Dataset for Training Machine Learning Models to Predict the Pathway Involvement of Metabolites

Cited by 7 publications

References 36 publications

Predicting The Pathway Involvement Of Metabolites Based on Combined Metabolite and Pathway Features

Predicting The Pathway Involvement Of Metabolites Based on Combined Metabolite and Pathway Features

Current approaches and outstanding challenges of functional annotation of metabolites: a comprehensive review

Predicting the Pathway Involvement of Metabolites in Both Pathway Categories and Individual Pathways

Contact Info

Product

Resources

About